migrate_sqlite_to_pgvector

Migrate a .swsearch SQLite index to a pgvector collection in PostgreSQL. Chunks are transferred in batches with their embeddings, metadata, and tags preserved.

Parameters

sqlite_path

strRequired

Path to the source .swsearch file.

connection_string

strRequired

PostgreSQL connection string (e.g., "postgresql://user:pass@host/db").

collection_name

strRequired

Name for the target pgvector collection.

overwrite

boolDefaults to false

Drop and recreate the collection if it already exists.

batch_size

intDefaults to 100

Number of chunks to insert per batch.

Returns

dict[str, Any] — Migration statistics containing:

source (str) — source file path
target (str) — target collection name
chunks_migrated (int) — number of successfully migrated chunks
errors (int) — number of failed chunks
config (dict) — source index configuration

Example

from signalwire.search import SearchIndexMigrator
migrator = SearchIndexMigrator(verbose=True)
stats = migrator.migrate_sqlite_to_pgvector(
    sqlite_path="./knowledge.swsearch",
    connection_string="postgresql://user:pass@localhost/mydb",
    collection_name="knowledge",
    overwrite=True,
)
print(f"Migrated {stats['chunks_migrated']} chunks with {stats['errors']} errors")